- Hi, I’m @amelie-iska
- I’m interested in deep learning applied to protein engineering and drug discovery and developing better reasoning agents with a universal multimodal token-modified transformer that will reason on any modality or any structured or unstructured data types with the generality of universal graph-to-graph function approximation, capability to reason with boolean circuits, the tropical ergodic dynamics of such models, and the implications on tranformer ASIC development and standardization of the industry to a single arch
- I’m currently thinking about composed architecture design and graph-of-thought reasoning in embedding space, with tropical attention and toric geometry methodologies, search methods with GFlowNets on GoT trajectories, etc.
- I’m looking to collaborate on AI for Proteins and Medicinal Design and agentic systems that are more expressive, generalizable to OOD data, and universal G2G function approximation
- How to reach me amelie.iska.math@gmail.com
- Pronouns: ...She/Her...
I'm working on AI applied to proteins and small molecules primarily for therapeutic applications in medicine and systems of composable neural networks
Pinned Loading
-
-
iska-net
iska-net PublicUniversal Graph Model training scaffold for reasoning on graph structured data with applications to biomedicine
TeX
-
Tropical_Quivers_of_Archs
Tropical_Quivers_of_Archs PublicA tropical, quiver representation theoretic take on composable learned functions with an embedding space native perspective
TeX 2
-
-
-
Flow-Matching
Flow-Matching PublicA differential topology introduction to flow-matching models with applications to generative biomolecular dynamics via quantum-accurate oracle training paradigms.
Something went wrong, please refresh the page to try again.
If the problem persists, check the GitHub status page or contact support.
If the problem persists, check the GitHub status page or contact support.